Landslide is one of the most common geological hazards, which causes a large number of property damage and loss of life in China every year. This case study was a 330-kV transmission line project located in Longnan City, Gansu Province, China, which is known as an area prone to landslides. A hybrid model of fractal theory-information value-random forests algorithm (FT-IV-RF) was proposed to evaluate the landslide susceptibility. First, sixteen landslide conditioning factors and pre-existing landslide events were selected as the initial evaluation indexes of landslide susceptibility from four datasets (geology, topography, climate and environment, and landslide inventory). Second, Pearson coefficient and sensitivity analyses were conducted to extract ten landslide conditioning factors with small correlation and large contribution to landslide occurrence from sixteen factors. Third, the weight of each class for a given factor were determined by using a combination of fractal theory and information value algorithms, which was regarded as one of input parameters and used to select the training samples in the random forest model. Four, k-means clustering was performed to classify the landslide susceptibility indices, which were predicted using the random forest model, into five levels to produce the landslide susceptibility map of the study area. Furthermore, the proposed model of FT-IV-RF model was validated by comparing with results obtained using information value (IV), back-propagation neural network (BPNN), and fuzzy logic (FL) models. Good agreements on the susceptibility estimation were observed among four models, and the hybrid model had the largest area under the curve (AUC) value of 0.996, indicating a good performance of the proposed hybrid model.